|Publication number||US7295995 B1|
|Application number||US 10/286,430|
|Publication date||Nov 13, 2007|
|Filing date||Oct 30, 2002|
|Priority date||Oct 30, 2001|
|Also published as||US7779014, US7953739, US20080040341, US20100293474|
|Publication number||10286430, 286430, US 7295995 B1, US 7295995B1, US-B1-7295995, US7295995 B1, US7295995B1|
|Inventors||Jeremy C. York, Robert Cottrell, Brent Smith|
|Original Assignee||A9.Com, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (18), Non-Patent Citations (6), Referenced by (96), Classifications (21), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application claims priority under 35 U.S.C. §119(e) from U.S. Provisional Application No. 60/336,235, filed Oct. 30, 2001, entitled “COMPUTER PROCESS FOR AUTOMATICALLY REFINING A LIST OF RECOMMENDED ITEMS BASED ON USER ACTIONS,” the entirety of which is hereby incorporated herein by reference.
1. Field of the Invention
This invention relates to data processing methods for adaptively refining the order and/or content of a list of items based upon the monitored actions of one or more users who access the list.
2. Description of the Related Art
One problem in the field of data processing and electronic catalogs is that of assisting users in efficiently locating items that are the most likely to be of interest. For example, when a web site displays a list of recommended items to a user, it would be desirable to present the list so as to include and emphasize those items that are the most likely to be of interest to the user given the particular context of the user.
One technique to address this problem involves presenting those items that, based on an analysis of browsing data, are or have recently been the most popular among a population of users. However, unless the user is able and willing to present feedback regarding the relevance of the presented items, these lists tend to be static and ineffective, particularly for users with specialized interests.
Another problem for online merchants is an inability to effectively target their marketing directly to those consumers who would be most interested in particular products. For example, existing product recommendation services tend to be effective only after an accurate profile of a particular consumer's interests has been developed. Thus, it may be difficult or impossible for a system that uses conventional recommendation algorithms to provide meaningful product recommendations to new customers or customers for which little information is known. Therefore, there is a continued need for ways to prioritize displays of items within an electronic catalog.
The invention comprises a system and method for ranking and presenting a list of items such that the items likely to be of greatest interest to a user viewing the list are presented first, or are otherwise displayed more prominently. The invention also includes methods for adaptively refining the contents of a list over time based on the actions of those who view the list. The invention is applicable to environments in which multiple users have access to a catalog or other repository of browsable items, including but not limited to products that are available for purchase.
In one particular embodiment, the lists are context-specific lists of recommended items (e.g., a list of suggested gifts for movie buffs). The orderings of such lists, and optionally the contents of the lists, are adaptively refined over time based on monitored actions performed by those who view the lists. In another embodiment, the lists are static-content lists, such as a list of all items falling within a particular item category.
To rank a list of items in accordance with the invention, a weight is preferably calculated for each item on the list, and these weights are used to rank the items for display. The item weights are preferably generated based on the monitored actions performed by users during or associated with viewing of the list. For example, in one embodiment, different permutations of each list are presented to different users and the monitored actions of these users with respect to the list items are used to calculate the item weights. The monitored actions may include, for example, item viewing events, “shopping cart add” events, purchase events, and/or any other type of action that reflects a user's preference for certain items over others.
By basing the weights of the list items on the activities of previous viewers of that same list, the weights used for ordering any given list may be separated from the weights used in ordering other lists (i.e., the same item may have different weights for different lists and contexts). Because the item weights for a given list are calculated based on user actions associated with the viewing of the lists, the weights strongly reflect user preferences within the browsing context associated with each list. For example, by calculating the item weights for a “gifts for graduates” list based solely on recorded user actions performed during viewing of this list, the list may be ordered based largely on the preferences of those who are actually shopping for a gift for a graduate.
By displaying only the most heavily weighted items within a given list, the invention may be used to adaptively refine a set of items presented to users. The contents (set of items) of a given list may also be adaptively modified over time by providing users an option to view items similar to those within the list, and based on user actions performed with respect to these “similar items,” automatically selecting specific similar items to add to the list. For example, if a relatively large number of those who, during viewing of a “gifts for graduates” list, select a “view similar items” option and then select item A, item A may automatically be added to the list.
A preferred embodiment of the invention will now be described with reference to the drawings summarized below. Throughout the drawings, reference numbers are re-used to indicate correspondence between reference elements.
A preferred embodiment will now be described and illustrated in which the system and associated methods are implemented using HTML (HyperText Markup Language) encoded web pages of a web site. As will be recognized, the invention may also be implemented using other types of multi-user interactive systems, including web sites that use other markup languages, online services networks that use proprietary data formats (e.g., America Online and MSN), and systems in which users browse and make purchases by voice.
Throughout the description, reference will be made to various implementation-specific details. These details are provided to fully illustrate a specific embodiment of the invention, and not to limit the scope of the invention. The various processes described herein are preferably performed by using software executed by one or more general-purpose computers. The processes could alternatively be embodied partially or entirely within special purpose hardware without altering the fundamental system described.
In particular, a “module” as used herein, may refer to any combination of software, firmware, or hardware used to perform the specified function or functions.
The modules described herein are preferably implemented as software modules, but may be represented partially or entirely in hardware or firmware. It is contemplated that the functions performed by these modules may also be embodied within either a greater or lesser number of modules than is described in the accompanying text. For instance, a single function may be carried out through the operation of multiple modules, or more than one function may be performed by the same module. The described modules may be implemented as hardware, software, firmware or any combination thereof. Additionally, the described modules may reside at different locations connected through a wired or wireless network, such as the Internet.
Although the items in the various embodiments described below are products, it will be recognized that the disclosed methods are also applicable to other types of items, such as documents, authors, musical artists, restaurants, chat rooms, other users, links to other web sites, web pages, or lists of items in a particular category.
The system and method described herein is designed to modify lists of items that are viewed by users of a website based on the context of each user at that website. “Context” is described in further detail herein, but may be considered to be the particular task the user is attempting to complete or a particular frame of mind of the user. By modifying lists of items to display items that are most likely to be of the interest to the user in the particular context of the user, the user is presented with more relevant items and is therefore more likely to make purchases at the website. A preferred method and system to develop and update such context driven lists is described herein.
The list 10 may also provide access to one or more additional pages that provide detail about each item 20 displayed. In
Although a list 10 may include any number of items 20, a single display page is usually constrained to display only a limited number of items at any one time. In
In addition to providing access to the items of a particular list, the list 10 may also include one or more links or buttons 45 that allow a user to view items that are similar or related to a selected item 20 on the list. Such ‘View Similar Items’ buttons 45 are shown for each item 20 displayed in the list 10. An option may also be provided (not shown) to view items that are collectively similar to all, or a selected subset, of the items in the list. Selecting the ‘View Similar Items’ button 45 for any particular item 20 or group of items will generate a new display page containing a different list of items (“similar items”). The list of similar items for a selected item 20, or for a selected group of items, may optionally be generated using the methods disclosed in U.S. Pat. No. 6,266,649 and published U.S. Application US 2002/0019763 A1, the disclosures of which are hereby incorporated by reference.
As noted above, it is desirable to order the items 20 displayed in the list 10 such that those items likely to be of the most interest to someone viewing the list appear earlier in the list. In general, by presenting items to users for which there is greater interest, more sales will be made through the web site. Presenting items of interest to users, however, is more than simply presenting the overall bestsellers to the user. For example, even though a particular item may be particularly popular among classical music aficionados, that item overall is likely to be much less popular than a new pop album by a current singing sensation. Furthermore, the classical music items that are most popular among classical music aficionados may not be the best selling items among classical music items overall. For instance, the general public may purchase a fairly large number of recordings of Beethoven's Fifth Symphony, while among classical music fans, recordings of Mozart's Requiem may outsell Beethoven's Fifth.
By knowing the context in which a visitor's activity occurs, it becomes possible to identify preferences associated with each context and present choices to users that are more likely to result in sales based upon each context. For example, if a user is operating in a context which is recognized to be that of a classical music fan, a recommendation of Mozart's Requiem will generally result in greater sales and related activity than a recommendation of Beethoven's Fifth Symphony. However, if a user is operating in the context of a general purchaser who is not a classical music fan, the Beethoven recommendation may generally result in better sales activity.
A user's context is typically that of the list currently being requested. For example, whenever a user requests a list associated with classical music, the user may be treated as falling within a “classical music fan context.” Thus, for example, the system may include many different browsable lists, each of which is statically assigned to a particular context.
As shown in
2. Generation of Activity and Exposure Data
When a web page is served to a user by the web server 120 and the user navigates through the website, two services are used to gather information about the user's visit to the website. First, a query log 140 and a query log analysis service 145 are used to generate an activity database 155 that contains a record of the user's interaction with the website. Second, a display tracking service 130 is used to generate an exposure database 135 that contains a record of how often an item is displayed in a particular context. Each service will be discussed in turn below.
As a user navigates the website, such as by traversing a catalog or sending requests for searches to the web server 120, the pages served to the user and the user's actions are stored in one or more query logs 140. The query logs are files or databases that contain a summary of the user's interaction with the website, for example, navigation decisions by the user, inquiries made by the user, links selected by the user, and additional information related to the activity of the user (e.g., dwell times on particular pages). Note that the activity that is stored in the query log may depend on the type of item viewed by the user and the path the user took to arrive at the particular item. For instance, if a user either searched for or traversed a catalog hierarchy to arrive at a specific author, then the activity that is stored in the query log may include both the path the user took to get to the author, as well as the activities of reading of reviews of the author's work, viewing detailed pages of the books by the author, or submitting a user review for a work by that author.
In a preferred embodiment, each record in the query log 140 preferably identifies the particular context, followed by a list of items viewed or otherwise acted upon by the user, as identified by the clicks, selections, or other user action performed in that context. Each query log record may also list some or all of the pages viewed during that user's session, in which case a look up table may be used to convert page identifiers to item or product identifiers. For purposes of this description, “context” refers generally to the particular task the user is attempting to complete or a particular frame of mind of the user. Contexts are typically, but not exclusively, defined by a series of one or more actions that a user takes before arriving at an ordered list, and/or by a name or other identifier of the list (e.g., “gifts for graduates”). For example, a user that answers a series of questions about a type of gift that they are searching for may arrive at a page similar to that depicted in
The query log 140 is periodically accessed by a query log analysis service 145. The query log analysis service 145 analyzes the activity that has taken place with respect to items in contexts of interest to the website operator. In a preferred embodiment, the query log analysis service analyzes one or more query logs 140 and generates an activity database 155 that contains a summary of activity data associated with each context and each item. In a preferred embodiment, the query log analysis service is a Perl script that is run periodically (e.g., once a day) on the list ordering system 100. Alternatively, the query log analysis service may rely on an underlying database to access the data.
The query log analysis service 145 scans the query logs 140 looking for one or more of the contexts listed in the context database 150. For each of the searched contexts from the context database that is found in the query log 140, the analysis service generates a list of the individual items associated with that context, as well as the activity associated with each of these items. This information is output to an activity database 155 that stores, for each context, the list of items and the activity associated with each item. Data associated with contexts not listed in the context database 150 are not analyzed and summarized by the query log analysis service 145.
It will be appreciated that items 210 a, 210 b . . . 210 n are not necessarily limited to products for sale. Items may also represent lists, authors, related web sites and other types of features or widgets that a user may show an interest in. While such items cannot generally be added to wish lists or shopping carts, other activities associated with such items may be used to indicate an interest of a user in the item in a particular context. For instance, if one item represents a list created by a particular user of her favorite books, a request by a user to mail the list to a friend may indicate the pertinence of the list to the particular context.
Although a particular data organization is shown in
In some embodiments it may be desirable to use multiple display tracking services 130 running on separate computing devices or servers. When this is the case, the individual data produced by the separate display tracking services may be combined into a single exposure database 135 at a variety of levels. For instance, each of the display tracking services 130 may load data into the same database 135 in order to produce a single database with the information from all of the various tracking services. In an alternative arrangement, the data from the separate display tracking services may be loaded into separate databases and aggregated at a later time. Moreover, the data from the display tracking services need not even be aggregated at all except when processed by the weighting service 160.
Although the term “viewing” is used herein to refer to the act of accessing product information, it should be understood that the user does not necessarily have to view the information about the product. Specifically, some merchants support the ability for users to browse their electronic catalogs by voice. For example, in some systems, users can access voiceXML versions of the site's web pages using a telephone connection to a voice recognition and synthesis system. In such systems, a user request for voice-based information about a product may be treated as a product viewing event.
3. Generation of Item Weights
Once the activity database 155 and exposure database 135 have been populated, the weighting service 160 may calculate a significance factor, or “weight,” to be assigned to a particular item in a particular context. The weighting service 160 periodically accesses the activity database 155, the exposure database 135, and the weighting database 165 in order to calculate such weights. As will be described in additional detail below, the weight of each item is used to select which items to display to users and the order of the items for display.
One preferred embodiment of the weighting service 160 for calculating a weight for an item within a particular context is shown in
The weighting service 160 begins by identifying the item that is to be analyzed at a step 310. At a step 320, a current exposure index (Ecurrent), is determined by accessing the exposure database 135 to determine how often the item has recently been displayed to users in the selected context. At a step 325, recent activity data associated with the item in the selected context is retrieved from the activity database 155. At a step 330, a current activity index (Acurrent) is calculated by summing the product of the quantity of each type of activity tracked and a factor associated with that type of activity. If, for example, the activities tracked for an item included the number of detail page views for the item (DV), the number of times the item was added to a wish list (WL), the number of times the item was added to a shopping cart (SC), and the number of times the item was actually purchased (PC), and if the activity factors applied to these four types of activities (DV, WL, SC, and PC) are respectively 1, 5, 10 and 15, then the current activity index for this context/item combination is calculated as follows:
A current=(1·DV)+(5·WL)+(10·SC)+(15·PC) (Eq. 1)
Individual activity factors are desirably assigned to each type of activity that is tracked by the system and used in the calculation of the current activity index. The particular values of the activity factors are assigned so as to approximate the relative importance of the corresponding activity as an indication of an interest level of the user in the item. For instance, in the example given, the activity factor for wish list adds (5) is five times greater than the activity factor for detail views (1), based on the assumption that adding an item to a wish list generally indicates a much stronger level of interest than merely viewing the detail page describing an item. The activity factors may be assigned by analyzing historical data, or may be assigned based upon market research or any other criteria known to those of skill in the art.
After the current exposure and activity indexes are determined, at a step 340 the weighting service 160 loads the historical exposure index (Ehistorical) and activity index (Ahistorical) from the weighting database 165. The weighting database contains the last-calculated value of the exposure index and the activity index for each context/item combination. In a preferred embodiment, the weighting database 165 comprises a flat file containing the context/item identifier and the associated exposure index and activity index. For new items, i.e., items that have no historical activity or exposure index associated with their display in a particular context, a technique is described in further detail below for providing a seeded value to ensure that the new item is ultimately exposed in lists.
Once the current and historical exposure and activity indices are calculated and loaded, new versions of each of the indices are calculated in accordance with the following equations:
A new =A current +αA historical (Eq. 2)
E new =E current +αE historical (Eq. 3)
where α (alpha) is a constant that represents the decay rate of historical data. Alpha is a value between 0 and 1 and is used to cause the historical data to be less significant in the calculation than the current data, thereby preventing early exposure and activity data from having undue significance over a period of time. In a preferred embodiment of the system, the value of alpha is approximately 0.99 to provide for a slow decay of the significance of historical values. For conditions in which a faster decay rate is desirable, a lower alpha value may be used. Faster decay rates are typically more appropriate for items that are seasonal in nature (for instance, holiday items), or for items that are known to be fashionable for only limited periods of time.
In an alternate embodiment, it may be desirable to use different alpha values for different context/item combinations, or for different categories of context/item combinations. For instance, a lower alpha value could be applied to all items in contexts that were known to represent seasonal or highly volatile markets. These individual alpha values could be stored in the weighting database 165 along with the historical exposure and activity indices for a given context/item combination if so desired. Those skilled in the art will recognize that alternate techniques to ensure that the effects of historical data are lessened over time may also be used.
Once the new exposure and activity indices have been calculated, at a step 345 the raw weight is calculated based upon the following equation:
The raw weight essentially represents the expected activity score that will occur when an item is displayed in a particular context. An adjustment factor must then be applied to the raw weight in order to prevent items that are only displayed a few times, but which were purchased by several of these users, from having unreasonably high weights. This circumstance may result in particular when a particular obscure item is searched for directly because a consumer knows he/she intends to purchase the item. At a block 350, an intermediate weight is therefore calculated by multiplying the raw weight by an adjustment factor as follows:
The use of the adjustment factor is based on the premise that activity on items having greater exposure is more reliable than activity on items having less exposure. Those skilled in the art will appreciate that other techniques may be used to create an adjustment factor having a similar effect.
After calculating the intermediate weight, at blocks 355 and 356 the historical exposure and activity indices are updated and stored in the weighting database 165. In the preferred embodiment of the weighting service, the new values of the exposure index (Enew) and activity index (Anew) are stored in the weighting database and used as the historical values in the next round of calculations. Those skilled in the art will recognize that a variety of techniques may be used to manipulate the new index values prior to storing as historical index values. Other possibilities include multiplying the new values by a constant prior to storing, or taking a weighted average between the new indices and the historical indices.
At a step 370, if there are additional items present in the activity database for the context being analyzed, the weighting service 160 returns to step 310 to select the next item and proceeds to repeat the described procedure for each of the identified items. When all of the items have been processed for a particular context and an intermediate weight has been determined for each of the items, at a step 380 the resulting intermediate weights are linearly redistributed to produce a final weight between 0 and 999 in accordance with the following equation:
where Wmax is equal to the largest intermediate weight for any item within the context. In this way, the range of final weights that the list generating component must process preferably falls within a consistent range regardless of the range of the actual weights produced, simplifying the processing required when generating lists in real time. At a step 385, the final weights are stored in a weight table 170 that correlates the context/item with the final weight.
In a preferred embodiment, the weighting service runs approximately once a day, preferably after completion of the analysis by the query log analysis service 145. In this way, new weights are generated for items in various contexts in response to the updated activity database 155. The weighting service may also be run more often if the activity database 155 is generated more frequently.
It will be appreciated that for new items added to a catalog, there will be no current or historical activity or exposure index associated with the display of the item in a particular context. When such information is lacking and a final weight cannot be calculated, an estimated final rank can be applied to the item. In one embodiment, a new item is assigned an estimated final weight of between 450 and 550 depending on the initial sales rank of the product as measured against other comparable products sold by the website operator (a higher sales rank pertaining to a weight of 550 and a lower sales rank pertaining to a weight of 450). Adding a new product with a final weight that falls in the middle of the weight range ensures that new products are not introduced into lists at a disproportionately high level, while at the same time ensuring that the item does not fail to be exposed on a list because it has a weight that is too low. Those skilled in the art will appreciate that numerous other techniques for introducing new items may also be applied, including varying the range or correlation with the sales rank, randomly assigning weights, assigning weights based on subjective item importance, and assigning weights in accordance with other website operator objectives.
4. List Ordering Module
The weights as determined by the weighting service 160 are used by the web server 120 to order lists of items for display in a user's browser 110. The web server 120 receives requests from a user's browser 110 or other system for accessing the contents of the vendor website, and in particular, a list of items. Before display to the user, the web server calls or otherwise executes a list ordering module 180 that orders the list that is to be displayed so that the most relevant items in the list are displayed to the user at the top of the list. The list ordering module takes as input the context in which the items are to be considered and a list of items that are to be ordered. The list ordering module returns an indication of the order in which the items should be listed for the particular context.
The preferred process for selecting, ordering, and displaying a list is shown schematically in
Initially, the web server 120 will receive a request for a display of items matching a particular request from a user. Such a request may, for instance, be for items related to a particular link the user has selected. The request for items may also be generated by requesting items that are similar to a list of items that are being viewed by the user. Those skilled in the art will recognize that any technique that generates a list of items, whether products or other items (such as authors or sub-lists) may be used to produce a list suitable for ordering based upon context. At a step 405, the user request determines the context that will be used to determine the list order. As discussed above, the context is recorded and tracked via the display tracking service 130 and query log analysis services 145 systems.
When the list request is received, at a step 410 the web server determines the content that matches the request and prepares a list of items that fit within the request. Typically, the list of items that is generated is greater than number of items that may be readily displayed on the browser display without scrolling. At a step 420, the context and the list of items to be displayed are passed to the list ordering module 180.
At a step 425, the list ordering module receives the context and the list of items and begins the necessary processing to generate an appropriate list order. At a step 430, the appropriate weights for the items are retrieved from the weight table 170 by matching the context and item identifiers with the appropriate entries in the weight table. While the weight table 170 need not be stored within the same machine as either the web server 120 or the list ordering module 180, it is necessary for the list ordering module to have access to the weight table in order to perform the appropriate ordering function.
Once the weights are retrieved for each of the items in the list, the items may be sorted. In a preferred embodiment, at a step 440 the items are sorted by the retrieved weights. Items that have no corresponding weights in the weight table (i.e., they have not yet appeared within this context) may be placed by default at a position in the middle of the list, or distributed somewhat randomly. At a step 445, some additional ordering of the list may optionally be performed. For example, one or more items may be randomly moved within the list to ensure that even seldom selected items get some exposure.
Those of skill in the art will also recognize that the list may be ordered using a combination of weighting and randomization. In an alternate embodiment, the ordering or weighting may also be used to eliminate those items which are weighted below a certain threshold value. Those of skill in the art will recognize that there are a variety of ways in which one may order or modify the list based at least in part upon the context sensitive weight of the items other than a purely weight-ordered list.
Once the list is ordered, at a step 450 the list items are returned to the web server in the order calculated. At a step 455, the web server may then apply one or more rules that further specify how the list should be displayed. For instance, the web server might be programmed to generate web pages listing only the first twenty items of lists with more than twenty entries, or may be programmed to filter out items the user has already purchased. Once the appropriate web page is generated, it is delivered to the user's browser for display at a step 460.
In addition to the preferred embodiment described above, there are alternate embodiments which make use of the same general configurations of the described components, but which may use different algorithms or techniques within some of the described components.
Another alternate embodiment of the system described above is substantially similar in overall structure to the system described, but makes use of a query tracking service rather than using a query log analysis service to summarize query logs. By using a query tracking service, the query log analysis service can be eliminated and processing may be performed substantially in real-time.
In one alternative embodiment, a query tracking service may comprise a module similar to the display tracking service described above. The query tracking service is a process which is used to track the activities being performed by various visitors to the web site. The activity of each user and the context in which that activity occurs is noted and recorded as it occurs in a database. This database may be substantially the same as the activity database that is produced by the query log analysis service as described above. As discussed with respect to the display tracking service above, the query tracking service may be implemented as a process which receives messages from the web server indicating the type of activity which has been performed by the user, as well as the context within which that activity was performed
Furthermore, as discussed above, it may be possible to have multiple query tracking services running simultaneously and feeding data to the same database of information. In this way, it is possible to have multiple web servers processing queries, but to have the results of activities performed placed into a single database or activity database for usage. As noted above, the data contained in such a database is substantially the same as the data which would appear in the activity database; however, rather than being produced by delayed analysis of data which was collected earlier, the activity database data may be maintained in a useful format in real-time or near real-time.
The functions performed by the query tracking service are preferably substantially similar to the functions performed in the creation of the query logs by the web servers and the summarizing of the query logs by the query log analysis service. However, by providing a tracking service which processes the incoming data in real time, the summaries are available substantially in real time, rather than being prepared only periodically by the query log analysis service. In this way, it is possible for the weighting function to be run at any time without the need for the query log analysis to be performed first. The ability to generate weights on a more frequent basis allows rapidly evolving trends to be captured and reflected in the web site's output more promptly.
The embodiment described above for the weighting service provides one technique for addressing the issue of generating appropriate levels of data for items without extensive display history. By using different algorithms based upon the total exposure of the context/item combination, items with low exposure can be provided with enough additional exposure to provide meaningful data to the weighting service and allow for more effective measurement of the interest in the item within that context. Those of skill in the art will recognize that it is also possible to alter the basis for the placement of low exposure items within the list in a variety of ways in order to generate meaningful exposure data. Furthermore, this exposure based system could be enabled only for a portion of the visitors to the web site in order to generate exposure data while still allowing most users to see the items in their calculated order.
Another alternate embodiment of the list ordering system described herein may make use of essentially the same steps, but distribute them differently within the various modules. For instance, certain preprocessing may be done in the query log analysis service 145 in order to simplify the later calculations performed by the weighting service 160. As an example, if the algorithm used in
In this way, less data need be stored within the activity database 155 prepared by the query log analysis service 145. Furthermore, the weighting service 160 may then simply read the activity index directly from the activity database 155 rather than performing calculations to derive the current activity index. Particularly if there is a low load on the hardware on which the query log analysis is run, and there is a need to minimize the processing load associated with the weighting service 160, such an alternate embodiment may be used to streamline the operation of the overall system. Those of skill in the art will recognize that a variety of such techniques may be used in order to best take advantage of the available processing power and latency requirements of the system, without altering the fundamental nature of the system described.
In other alternate preferred embodiments of the described systems and techniques, different values may be used for the constant alpha within the equations above. Because this constant represents effects that are modeled based upon real world conditions (e.g. the rate at which interest decays in currently popular items, the amount of historical significance past popularity has in predicting future popularity, etc.), these values may occasionally require adjustment. In additional to periodic reevaluation of such factors in order to tune the behavior of the ordering system, in alternate embodiments this value may be assigned based upon certain objective factors, such as the type of item with which the factor is associated (e.g. book, music or video), or the cost of the item. These and such other adjustments as are known to those of skill in the art will allow for more effective ordering of the lists by the system.
In addition to the general ordering techniques described above, the list ordering module 180 may also be configured to perform additional filtering or ordering calculations based upon other information. For instance, in one alternate embodiment, the list ordering module also has access to an inventory database that provides an indication of which items are currently in stock and what the approximate lead time is for items that are out of stock. The inventory information is used to prioritize certain lists based not only on the likelihood that a user will want to purchase an item, but also based upon the availability of the item. In this way, items that are available for immediate shipment may appear higher on the list.
The query logs may also include, for each list viewing event, information about whether the user selected the ‘View Similar Items’ link with respect to any of the items in the list. If a ‘View Similar Items’ link was selected, the query log also preferably includes information about any actions (e.g., item viewing events, shopping cart adds, wish list adds, purchases, etc.) performed by the user with respect to the displayed similar items. Through the use of the ‘View Similar Items’ function, display and activity tracking data may be collected in various contexts for the similar items that were not part of the list as it was initially populated. For instance, a list of “Gifts for Graduates” may be initially populated with items that are manually selected by an online merchant, and/or may be populated automatically based on a seeding algorithm. Over time, in addition to data being collected about the items already in the list, context, exposure and activity data may also be collected for those similar items that appear in subsequent lists produced in response to users choosing to ‘View Similar Items’ for various items on the “Gifts for Graduates” list. For example, the merchant might have initially included the book “Everything I Needed to Know I Learned in Kindergarten” in the “Gifts for Graduates” list. As this list is presented to users over time, some users may opt to ‘View Similar Items’ to this particular book. A variety of similar items (typically other books) will be displayed in response to this selection, and some users may perform one or more types of activities with respect to the similar items—such as by viewing, purchasing, or adding one or more of the similar items to a wish list. Each of these activities is tracked and recorded in the activity database.
Once the activity tracking data for similar items has been collected, the data may be used to automatically refine the contents of the list. Even though the similar items typically are not part of the initial “Gifts for Graduates” list, they may be added to the list as the result of activity and exposure data collected through this process. As time goes by and the weighting service is run to generate appropriate weights to use in ordering the items in the list, weights may be generated for similar items that were not part of the original list. The list ordering service can use the weights of the similar items to determine whether or not to add any of the similar items to the list. For example, once the exposure level of a particular similar item exceeds a selected threshold value, the similar item may be added to the list based upon its weight. In this way, new items that may not have been initially recognized as being likely to generate interest within a particular context may be added to the list of items displayed to a user in that context.
6. Example Applications
The above techniques and systems may be useful in a variety of applications in which lists of items, such as products, are presented to a user based upon differing contexts. Example applications that make use of such techniques and systems are described below.
One particular embodiment is a “gift wizard” system in which each of a plurality of browsable gift-context-specific lists evolve over time (i.e., change in content) to reflect the actions of those who view that list. One example of a gift-context-specific list is a “Gifts for Graduates” list, as mentioned above with reference to
Once an initial list is created and given a user-viewable gift context name (such as “Gifts for Movie Buffs”), the display tracking service and query logs may be used as described above to develop the appropriate databases of exposure and activity information associated with each item displayed within this context. As addressed above, data associated with additional items which were not originally included in the list may also be tracked when such items are displayed in response to users selecting to ‘View Similar Items’ to any particular item displayed in the list.
When the weighting service is run, weights for each of the displayed items are produced based upon the activity of the various users who viewed the list. As discussed above, this may include weights for items not originally included in the list when it was set up by the merchant. Based upon these weights and the threshold values chosen for use in the weighting and list ordering algorithms, the list of items the is displayed to the user when viewing items within the “Gifts for Graduates” context may evolve over time. Thus, the item weights may be used both to select the items to displayed to users, and to determine the order in which the displayed items are to be presented.
Such evolution may be advantageous in allowing additional items not originally recognized as having particular value within the context to be brought to the attention of users. Similarly, items that were initially included on the list but have been performing relatively poorly in terms of the tracked user activity will naturally tend to move down the list. By placing an upper limit on the number of items to be included in any particular display of the list, items may effectively be removed from the list if they are weighted below non-original items automatically added to the list.
Note that although the application above is described in terms of a “gift wizard,” such an application may be used beyond the confines of gifts. For instance, such an evolving list of items may be used for specialty stores related to particular genres of music, film or literature, as well for broad categories of interest that are not easily identified.
Another application that can make use of the techniques and systems described herein makes use of a list which may desirably remain static, unless explicitly changed by the operator of the system. For example, online sales systems commonly include a “browse tree” through which users can browse the various catalog items (products) based on predefined item categories and subcategories. See, e.g., U.S. Pat. No. 6,466,918, the disclosure of which is hereby incorporated by reference. A given category within the tree in the form of a browse node that includes a list of all items falling within that category. In accordance with the invention, the display order (but not the content) of each such list is modified over time to reflect the actions of those who view the list. Thus, for example, the display order of the list associated with the browse node “sports\olympics\videos” may change over time based on user actions performed during viewing of this list, without regard to other user actions performed on the items within this list.
Such a system may be useful when a list is categorically based. For instance, a list which includes all country music within the population of available items might be kept without allowing for the possibility of the list itself evolving over time. This is because users who are browsing a list entitled “Country music” will generally not expect to find items on the list other than country music items. Such arrangements may also be useful when a list is limited to a particular category which only changes slowly over time. For example, the collection of Star Trek books might be handled using such a static browse node application. Although there are hundreds of books in this category by a variety of authors, there is no need to let this list change except for when a new Star Trek book is released.
The various embodiments described above in accordance with the present invention thus provide a system and technique to provide a prioritized order to a list of items for which user activity data and context have been tracked. By using this data to estimate the level of interest in the items within the various display contexts, a list may be ordered, and optionally modified in content, so as to emphasize items which are most relevant in the list's context.
It is to be understood that not necessarily all objects or advantages may be achieved in accordance with any particular embodiment described herein. Thus, for example, those skilled in the art will recognize that certain embodiments may be configured to operate in a manner that achieves or optimizes one advantage or group of advantages as taught herein without necessarily achieving other objects or advantages as may be taught or suggested herein.
Furthermore, the skilled artisan will recognize the interchangeability of various features from different embodiments. In addition to the variations described herein, other known equivalents for each feature can be mixed and matched by one of ordinary skill in this art to construct list ordering or prioritizing systems in accordance with principles of the described system.
Although the systems and techniques above have been disclosed in the context of certain preferred embodiments and examples, it will be understood by those skilled in the art that these techniques and systems extend beyond the specifically disclosed embodiments to other alternative embodiments and/or uses, including obvious modifications and equivalents thereof.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5446891||Nov 2, 1994||Aug 29, 1995||International Business Machines Corporation||System for adjusting hypertext links with weighed user goals and activities|
|US5749081||Apr 6, 1995||May 5, 1998||Firefly Network, Inc.||System and method for recommending items to a user|
|US5825851||Jun 17, 1996||Oct 20, 1998||Samsung Electronics Co., Ltd.||Method and apparatus for automatically transferring a caller to a selected extension|
|US5848396||Apr 26, 1996||Dec 8, 1998||Freedom Of Information, Inc.||Method and apparatus for determining behavioral profile of a computer user|
|US6085229 *||May 14, 1998||Jul 4, 2000||Belarc, Inc.||System and method for providing client side personalization of content of web pages and the like|
|US6256633 *||Jun 25, 1998||Jul 3, 2001||U.S. Philips Corporation||Context-based and user-profile driven information retrieval|
|US6334127 *||Jul 17, 1998||Dec 25, 2001||Net Perceptions, Inc.||System, method and article of manufacture for making serendipity-weighted recommendations to a user|
|US6438579 *||Jul 14, 2000||Aug 20, 2002||Agent Arts, Inc.||Automated content and collaboration-based system and methods for determining and providing content recommendations|
|US6466970||Jan 27, 1999||Oct 15, 2002||International Business Machines Corporation||System and method for collecting and analyzing information about content requested in a network (World Wide Web) environment|
|US6493702 *||May 5, 1999||Dec 10, 2002||Xerox Corporation||System and method for searching and recommending documents in a collection using share bookmarks|
|US6606102||Jun 2, 2000||Aug 12, 2003||Gary Odom||Optimizing interest potential|
|US6901378 *||Mar 2, 2000||May 31, 2005||Corbis Corporation||Method and system for automatically displaying an image and a product in a page based on contextual interaction and metadata|
|US7031961 *||Dec 4, 2000||Apr 18, 2006||Google, Inc.||System and method for searching and recommending objects from a categorically organized information repository|
|US7089237 *||Jan 26, 2001||Aug 8, 2006||Google, Inc.||Interface and system for providing persistent contextual relevance for commerce activities in a networked environment|
|US20010014868||Jul 22, 1998||Aug 16, 2001||Frederick Herz||System for the automatic determination of customized prices and promotions|
|WO1995029451A1 *||Apr 24, 1995||Nov 2, 1995||Apple Computer, Inc.||System for ranking the relevance of information objects accessed by computer users|
|WO1999041694A1 *||Feb 12, 1999||Aug 19, 1999||Yahoo, Inc.||Search engine using sales and revenue to weight search results|
|WO2000017792A1||Sep 10, 1999||Mar 30, 2000||Amazon.Com, Inc.||Collaborative recommendations using item-to-item similarity mappings|
|1||B. Mobasher, R. Cooley, and J. Srivastava, "Automatic Personalization Based on Web Usage Mining," Communications of the ACM, vol. 43, Issue.8, pp. 142-151, dated Aug. 2000.|
|2||D. Widyantoro, T. Loerger and J. Yen, "An Adaptive Algorithm for Learning Changes in User Interests," Proceedings of the Eighth International Conference on Information Knowledge Management, pp. 405-412, ACM Press 1999.|
|3||J. Andersen, A. Giversen, A. Jensen, R. Larsen. T. Pederson and J. Skyt, "Analyzing Clickstreams Using Subsessions," Proceedings of the Third ACM International Workshop on Data Warehousing and OLAP, pp. 25-32, ACM Press 2000.|
|4||K. Yu, X. Xu, M. Ester and H. Kriegel, "Selecting Relevant Instances for Efficient and Accurate Collaborative Filtering," Proceedings of the Tenth International Conference on Conference on Information and Knowledge Management, pp. 239-246, ACM Press 2001.|
|5||R. Kohavi, "Mining E-Commerce Data: The Good, the Bad, and the Ugly," Proceedings of the Seventh ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 8-13, ACM Press 2001.|
|6||S. Middleton, D. De Roure and N. Shadbolt, "Capturing Knowledge of User Preferences: Ontologies in Recommender Systems," Proceedings of the International Conference on Knowledge Capture, pp. 100-107, ACM Press 2001.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7594189||Mar 29, 2006||Sep 22, 2009||Amazon Technologies, Inc.||Systems and methods for statistically selecting content items to be used in a dynamically-generated display|
|US7698297 *||Apr 13, 2010||Apple Inc.||Accessing digital media|
|US7702541 *||Dec 15, 2000||Apr 20, 2010||Yahoo! Inc.||Targeted e-commerce system|
|US7739295 *||Jun 20, 2003||Jun 15, 2010||Amazon Technologies, Inc.||Method and system for identifying information relevant to content|
|US7779014||Oct 19, 2007||Aug 17, 2010||A9.Com, Inc.||Computer processes for adaptively selecting and/or ranking items for display in particular contexts|
|US7831572 *||Nov 9, 2010||Microsoft Corporation||Maintaining and displaying user ordered lists|
|US7953739||May 31, 2011||A9.Com, Inc.||Automated discovery of items associated with a context based on monitored actions of users|
|US7974964 *||Jan 17, 2007||Jul 5, 2011||Microsoft Corporation||Context based search and document retrieval|
|US7979321||Jul 12, 2011||Ebay Inc.||Merchandising items of topical interest|
|US7979445 *||Oct 7, 2010||Jul 12, 2011||Amazon Technologies, Inc.||Processes for assessing user affinities for particular item categories of a hierarchical browse structure|
|US8019650||Sep 13, 2011||Amazon Technologies, Inc.||Method and system for producing item comparisons|
|US8060466||Nov 15, 2011||Amazon Technologies, Inc.||Service for accepting and selectively exposing user-created lists of items|
|US8090621 *||Jun 27, 2007||Jan 3, 2012||Amazon Technologies, Inc.||Method and system for associating feedback with recommendation rules|
|US8095523 *||Jan 10, 2012||Baynote, Inc.||Method and apparatus for context-based content recommendation|
|US8117060||Dec 20, 2007||Feb 14, 2012||Ebay Inc.||Geographic demand distribution and forecast|
|US8121905||May 31, 2011||Feb 21, 2012||Ebay Inc.||Merchandising items of topical interest|
|US8131722 *||Jun 29, 2007||Mar 6, 2012||Ebay Inc.||Search clustering|
|US8166155||Apr 24, 2012||Amazon Technologies, Inc.||System and method for website experimentation|
|US8185608||May 22, 2012||Amazon Technologies, Inc.||Continuous usability trial for a website|
|US8190603 *||May 29, 2012||Fujitsu Limited||Information providing system, information providing method and information providing record medium|
|US8219447||Jul 10, 2012||Amazon Technologies, Inc.||Real-time adaptive probabilistic selection of messages|
|US8271357 *||Sep 18, 2012||Ebay Inc.||Presenting items based on activity rates|
|US8301512||Oct 30, 2012||Ebay Inc.||Product identification using multiple services|
|US8301623||Oct 30, 2012||Amazon Technologies, Inc.||Probabilistic recommendation system|
|US8402068||Dec 7, 2000||Mar 19, 2013||Half.Com, Inc.||System and method for collecting, associating, normalizing and presenting product and vendor information on a distributed network|
|US8429087||Sep 9, 2010||Apr 23, 2013||Alibaba Group Holding Limited||System and method of optimal time for product launch an withdraw in e-commerce|
|US8452727||Aug 1, 2011||May 28, 2013||Amazon Technologies, Inc.||Service for accepting and selectively exposing user-generated lists|
|US8478664||Oct 25, 2011||Jul 2, 2013||Amazon Technologies, Inc.||Recommendation system with user interface for exposing downstream effects of particular rating actions|
|US8554641||Jan 10, 2012||Oct 8, 2013||Ebay Inc.||Merchandising items of topical interest|
|US8577753 *||Oct 22, 2008||Nov 5, 2013||Amazon Technologies, Inc.||Community-based shopping profiles|
|US8589398 *||Feb 3, 2012||Nov 19, 2013||Ebay Inc.||Search clustering|
|US8595084||Sep 14, 2012||Nov 26, 2013||Ebay Inc.||Presenting items based on activity rates|
|US8645221 *||Jun 16, 2010||Feb 4, 2014||Amazon Technologies, Inc.||Ranking of items as a function of virtual shopping cart activity|
|US8655730||Sep 28, 2011||Feb 18, 2014||Amazon Technologies, Inc.||Selecting advertisements based on advertising revenue model|
|US8732528||Jan 6, 2012||May 20, 2014||Amazon Technologies, Inc.||Measuring test effects using adjusted outlier data|
|US8751470 *||May 31, 2012||Jun 10, 2014||Google Inc.||Context sensitive ranking|
|US8768937||Mar 11, 2013||Jul 1, 2014||Ebay Inc.||System and method for retrieving and normalizing product information|
|US8782074||Mar 22, 2010||Jul 15, 2014||Amazon Technologies, Inc.||Method and system for identifying information relevant to content|
|US8818977||Jul 12, 2012||Aug 26, 2014||Google Inc.||Context sensitive ranking|
|US8819039 *||Aug 25, 2003||Aug 26, 2014||Ebay Inc.||Method and system to generate a listing in a network-based commerce system|
|US8849807||May 24, 2011||Sep 30, 2014||Mark F. McLellan||Active search results page ranking technology|
|US8862574 *||Apr 30, 2009||Oct 14, 2014||Microsoft Corporation||Providing a search-result filters toolbar|
|US8863002||Oct 4, 2010||Oct 14, 2014||Ebay Inc.||Method and system to dynamically browse data items|
|US8892506||Feb 11, 2013||Nov 18, 2014||Amazon Technologies, Inc.||Service for accepting and selectively exposing user-generated lists|
|US8949712 *||Feb 6, 2013||Feb 3, 2015||Amazon Technologies, Inc.||Assigning slots to content in a pipeline|
|US8965998||Mar 19, 2003||Feb 24, 2015||Amazon Technologies, Inc.||Adaptive learning methods for selecting web page components for inclusion in web pages|
|US9037637||Feb 6, 2012||May 19, 2015||J.D. Power And Associates||Dual blind method and system for attributing activity to a user|
|US9043351 *||Jan 30, 2013||May 26, 2015||A9.Com, Inc.||Determining search query specificity|
|US9076173||Oct 29, 2012||Jul 7, 2015||Ebay Inc.||Product identification using multiple services|
|US9076179||Jun 21, 2013||Jul 7, 2015||Amazon Technologies, Inc.||Recommendation system with user interface for exposing downstream effects of paricular rating actions|
|US9135359||Jan 16, 2015||Sep 15, 2015||Amazon Technologies, Inc.||Adaptive learning methods for selecting page components to include on dynamically generated pages|
|US9152624 *||Nov 23, 2010||Oct 6, 2015||Retail Optimization International, Inc.||Systems and methods for visual presentation and navigation of content using data-based image analysis|
|US9171056||Jun 16, 2014||Oct 27, 2015||Paypal, Inc.||System and method for retrieving and normalizing product information|
|US9245294 *||Sep 14, 2012||Jan 26, 2016||Amazon Technologies, Inc.||Providing separate views for items|
|US9286271||Apr 30, 2013||Mar 15, 2016||Google Inc.||Providing an electronic document collection|
|US9292479||Apr 30, 2013||Mar 22, 2016||Google Inc.||Providing an electronic document collection|
|US9323849 *||Jun 16, 2011||Apr 26, 2016||Amdocs Software Systems Limited||Personalising content provided to a user|
|US20020016741 *||Dec 15, 2000||Feb 7, 2002||Black Peter M.||Targeted e-commerce system|
|US20020107861 *||Dec 7, 2000||Aug 8, 2002||Kerry Clendinning||System and method for collecting, associating, normalizing and presenting product and vendor information on a distributed network|
|US20040128320 *||Aug 25, 2003||Jul 1, 2004||Steve Grove||Method and system to generate a listing in a network-based commerce system|
|US20040215611 *||Mar 12, 2004||Oct 28, 2004||Apple Computer, Inc.||Accessing media across networks|
|US20060212362 *||Feb 13, 2006||Sep 21, 2006||Donsbach Aaron M||Method and system for producing item comparisons|
|US20070118441 *||Jun 28, 2006||May 24, 2007||Robert Chatwani||Editable electronic catalogs|
|US20080040341 *||Oct 19, 2007||Feb 14, 2008||York Jeremy C||Computer processes for adaptively selecting and/or ranking items for display in particular contexts|
|US20080060257 *||Sep 8, 2006||Mar 13, 2008||The Technology Store, Inc.||Biodiesel production and use in oil sands processing|
|US20080120292 *||Jun 29, 2007||May 22, 2008||Neelakantan Sundaresan||Search clustering|
|US20080172364 *||Jan 17, 2007||Jul 17, 2008||Microsoft Corporation||Context based search and document retrieval|
|US20090030803 *||Jul 25, 2007||Jan 29, 2009||Sunil Mohan||Merchandising items of topical interest|
|US20090037355 *||Aug 8, 2008||Feb 5, 2009||Scott Brave||Method and Apparatus for Context-Based Content Recommendation|
|US20090083259 *||Aug 22, 2008||Mar 26, 2009||Fujitsu Limited||Information providing system, information providing method and information providing record medium|
|US20090138472 *||Apr 3, 2006||May 28, 2009||Wine Science Ltd.||Method of Supplying Information Articles at a Website and a System for Supplying Such Articles|
|US20090150214 *||Dec 11, 2007||Jun 11, 2009||Sunil Mohan||Interest level detection and processing|
|US20090164294 *||Dec 20, 2007||Jun 25, 2009||Ebay Inc.||Geographic demand distribution and forecast|
|US20090210322 *||Feb 12, 2009||Aug 20, 2009||Karen Elizabeth Gould Stark||Method for creating custom blended cosmetics|
|US20090234875 *||Mar 13, 2008||Sep 17, 2009||International Business Machines Corporation||System and methods for providing product metrics|
|US20090259620 *||Apr 11, 2008||Oct 15, 2009||Ahene Nii A||Method and system for real-time data searches|
|US20090307047 *||Dec 10, 2009||International Business Machines Corporation||System to Defer and Refer Advertisements Presented in Conjunction with Media Device Control Interfaces|
|US20100070342 *||Aug 31, 2009||Mar 18, 2010||Ebay Inc.||Regional demand and supply comparison|
|US20100169160 *||Dec 30, 2008||Jul 1, 2010||Ebay Inc.||Gift recommendation method and system|
|US20100281417 *||Nov 4, 2010||Microsoft Corporation||Providing a search-result filters toolbar|
|US20100293474 *||Jul 26, 2010||Nov 18, 2010||York Jeremy C||Automated discovery of items associated with a context based on monitored actions of users|
|US20110022606 *||Oct 7, 2010||Jan 27, 2011||Mason Zachary J||Processes for assessing user affinities for particular item categories of a hierarchical browse structure|
|US20110076663 *||Apr 28, 2010||Mar 31, 2011||Retail Optimization International||Systems and methods for selecting survey questions and available responses|
|US20110093494 *||Oct 4, 2010||Apr 21, 2011||Ebay Inc.||Method and system to dynamically browse data items|
|US20110099085 *||Oct 23, 2009||Apr 28, 2011||Scot Fraser Hamilton||Product identification using multiple services|
|US20110252297 *||Oct 13, 2011||Amdocs Software Systems Limited||Personalising content provided to a user|
|US20120185446 *||Feb 3, 2012||Jul 19, 2012||Neelakantan Sundaresan||Search clustering|
|US20120290936 *||Jul 30, 2012||Nov 15, 2012||International Business Machines Corporation||Automatic generation of task scripts from web browsing interaction history|
|US20130211964 *||Jan 31, 2013||Aug 15, 2013||Global Village Concerns||Systems and methods for generation of an online store|
|US20130218717 *||Jan 31, 2013||Aug 22, 2013||Global Village Concerns||Systems and methods for generation of a registry in an online store|
|US20130226730 *||Apr 2, 2013||Aug 29, 2013||Target Brands, Inc.||Gift registry graphical user interface|
|USRE45793 *||Apr 12, 2012||Nov 3, 2015||Apple Inc.||Accessing digital media|
|EP2533163A1 *||Feb 3, 2011||Dec 12, 2012||eBay Inc.||List display on the basis of list activities and related applications|
|WO2011095667A1 *||Feb 3, 2011||Aug 11, 2011||Ebay, Inc.||List display on the basis of list activities and related applications|
|WO2012148315A2 *||Apr 26, 2012||Nov 1, 2012||Rawllin International Inc.||Generating product recommendations based on dynamic product context data and/or social activity data related to a product|
|WO2012148315A3 *||Apr 26, 2012||Apr 11, 2013||Rawllin International Inc.||The systems and methods are described for generating online product recommendations|
|U.S. Classification||705/26.8, 707/999.001, 707/999.005, 707/999.003, 707/999.004, 707/999.01, 707/999.002, 705/27.1|
|Cooperative Classification||Y10S707/99932, Y10S707/948, Y10S707/99931, Y10S707/99933, Y10S707/99934, Y10S707/99935, G06Q30/0641, G06Q30/0633, G06Q30/02|
|European Classification||G06Q30/02, G06Q30/0633, G06Q30/0641|
|Feb 10, 2003||AS||Assignment|
Owner name: AMAZON.COM, INC., WASHINGTON
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:YORK, JEREMY C.;COTTRELL, ROBERT;SMITH, BRENT;REEL/FRAME:013728/0630
Effective date: 20030203
|Sep 24, 2004||AS||Assignment|
Owner name: A9.COM, INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:AMAZON.COM, INC.;REEL/FRAME:015173/0742
Effective date: 20040810
|Nov 8, 2004||AS||Assignment|
Owner name: A9.COM, INC., CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:AMAZON.COM, INC.;REEL/FRAME:015341/0120
Effective date: 20040810
Owner name: A9.COM, INC.,CALIFORNIA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:AMAZON.COM, INC.;REEL/FRAME:015341/0120
Effective date: 20040810
|May 9, 2011||FPAY||Fee payment|
Year of fee payment: 4
|May 13, 2015||FPAY||Fee payment|
Year of fee payment: 8